Formulary decision on a new medicine at the margin of cost effectiveness

An NHS integrated care board's medicines committee decides whether to list a new medicine, restrict it to a subgroup, defer pending a better price, or decline. The log states the displaced health expl...

Published content

1. node: Context. The organisation is an NHS integrated care board in England, funding care within a fixed annual allocation. The medicine treats a chronic condition affecting approximately 900 of its patients, and offers a modest improvement in quality of life over current standard care with no demonstrated survival benefit. The manufacturer has applied for local listing with an economic model and a confidential discount. No NICE appraisal of the medicine is scheduled and specialists are asking for access now, so the committee has to decide without national guidance.
2. node: Should the medicines committee add the new medicine to the formulary without restriction, add it restricted to a defined subgroup, decline pending a revised commercial offer, or decline to list it?
3. node: Rationale. Unrestricted listing has the strongest evidence base of the options, because the submitted analysis is complete and clinically plausible for exactly this population.
4. node: Rationale. Restricting to the more severe subgroup directs the medicine to the patients in whom the modelled benefit is largest, which is the response the appraisal methods contemplate when a technology is cost effective for part of a licensed population and not for all of it.
5. node: Rationale. The obstacle to listing is the price rather than the evidence, and stating the ratio a revised offer would need to reach makes the decision reviewable and gives the manufacturer something specific to respond to.
6. node: Rationale. Declining is the clearest statement of the committee's position and does not hold out a prospect the committee cannot deliver.
7. node: Decision-maker. The medicines committee, chaired by the Medical Director, deciding collectively with the decision and its reasons minuted. The Chief Pharmacist presented the appraisal and does not vote.
8. node: Scope. The committee decides whether and on what terms the medicine is available to be prescribed here. It does not set the price, licence the medicine, or determine national guidance, and it cannot decline a medicine that NICE has recommended in a technology appraisal.
9. node: Rationale. Deferral preserves every option. Listing later is possible; withdrawing later, once patients are established, is largely not.
10. node: Consulted. The specialty clinicians who would prescribe, on the subgroup definition and whether it could be applied at the point of prescribing; the finance director on displacement within the medicines budget; and the patient representative on the committee, who recorded that patients waiting are visible while patients displaced are not, and asked that the minute say so.
11. node: Risk. At a ratio above the threshold the system loses more health than the medicine provides. The patients who lose it are not identifiable and will never know, which is what makes this the easiest harm in the decision to discount.
12. node: Risk. A subgroup defined after the trial, which was not powered for it, will usually produce a ratio that looks acceptable. The committee would be relying on the least robust number in the submission to justify the option that appears most reasonable.
13. node: Risk. It forecloses a price negotiation that the deferral option keeps open, at no benefit: the two options have identical figures on every criterion, and what separates them is that the deferral states what would change the answer.
14. node: Action. Communicate to the manufacturer the ratio a revised offer would need to reach, and the subgroup evidence that would be required if a restricted listing were to be considered instead. A deferral that does not say what would change the answer is a decline in slower form.
15. node: Constraint. The budget is fixed. Money spent on this medicine is not raised from outside the system; it displaces other activity, and the health forgone in that displaced activity is a consequence of this decision whether or not it is measured.
16. node: Risk. Once 900 patients are established on the medicine the decision is effectively permanent, and a later price rise or evidence revision would be met from a position with no practical option to withdraw.
17. node: Risk. Two previous restricted listings drifted beyond their criteria within two years with no audit in place to detect it. If prescribing drifts here, the organisation has adopted the unrestricted option without deciding to, and at the unrestricted option's net health effect.
18. node: Risk. Patients who would benefit are not treated while the position is held, and the wait is indefinite because the committee does not control whether an offer comes. That cost falls on identifiable patients, unlike the displaced health, which is why it will be argued more forcefully.
19. node: Risk. A decline recorded without stating what would change the answer is harder to defend to patients and clinicians than one that does, and invites the reading that the committee decided on cost alone.
20. node: Action. Establish a prescribing audit for restricted listings before the next restricted decision of any kind. The reason the restricted option cannot be recommended here is an absence of assurance rather than an absence of merit, and that absence is fixable.
21. node: Constraint. A decision to list must be applied consistently to every patient meeting the stated criteria. The committee cannot list on terms it is unable to define, because the definition is what makes the decision applicable rather than discretionary.
22. node: Action. Record in the minute what was weighed besides the ratio, including the position put by the patient representative. The methods expect deliberation rather than computation, and a minute that records only the arithmetic does not evidence that deliberation took place.
23. node: Guidance. The NICE health technology evaluations manual sets out the reference case for economic evaluation, in which the incremental cost-effectiveness ratio is compared against a threshold range, stated in the manual as £20,000 to £30,000 per QALY. The government announced in late 2025 that the range would rise to £25,000 to £35,000, and this committee applies the revised range. The ratio is described as the starting point for deliberation rather than as determining the outcome, with other considerations including judgements of social value also bearing on it.
24. node: Research. Work by Claxton and colleagues on the estimation of the NICE cost-effectiveness threshold describes the threshold as an estimate of health forgone as other activity is displaced to accommodate the additional cost of a new technology. On that account, a ratio above the threshold means the system loses more health than the medicine delivers. The net health effect criterion below states that directly, at £25,000 per QALY, rather than leaving it implicit in a ratio.
25. node: Review trigger. Reconsider immediately on receipt of a revised offer bringing the full-population ratio to £25,000 per QALY or lower. The committee's position is about price and not about the medicine.
26. node: Research. An analysis of NICE appraisal policy from 1999 to 2018 records that the ratio and the threshold have been the consistent tools of decision-making, with some allowance for social and ethical values, and that decisions are not determined by the ratio alone. A committee is therefore expected to deliberate rather than to compute, and to record what it weighed when it departed from the arithmetic.
27. node: Review trigger. Reconsider the restricted option if the manufacturer supplies a probabilistic analysis at subgroup level, and once the prescribing audit is operating. Both deficiencies are remediable and neither is a judgement about the medicine.
28. node: Guidance. NICE methods have required probabilistic sensitivity analysis in submitted appraisals since 2004, and parameter values used in base case and sensitivity analyses are expected to be clinically plausible and consistent with the data. The submitted analysis meets this requirement for the whole licensed population but is sparse for the subgroup, which bears on the certainty recorded for the restricted option.
29. node: Review trigger. If NICE recommends the medicine in a technology appraisal, the organisation must make it available within the statutory period and this decision lapses.
30. node: Assumption. The threshold range used here represents the health this system actually forgoes when it displaces activity. It is taken from national work rather than estimated locally, and this organisation's own displacement may be more or less efficient than the national estimate assumes. The NICE range is used here as the measure of displaced health. Empirical estimates of the health actually displaced in the NHS, such as Claxton and colleagues' figure of about £13,000 per QALY, are lower, and on them every listing option would reduce net health.
31. node: Assumption. Prescribing will follow the restriction if one is set. Two previous restricted listings drifted materially beyond their stated criteria within two years, and no audit was in place to detect it at the time.
32. node: List without restriction, for the full licensed population
33. node: List restricted to the subgroup with more severe disease, where the modelled benefit is larger
34. node: Decline to list pending a revised commercial offer, with the ratio the offer would need to reach stated
35. node: Decline to list
36. node: Recommendation. That the committee decline to list pending a revised commercial offer, and state that the offer must bring the full-population ratio to no more than £25,000 per QALY, the point at which the net health effect is no longer negative at the threshold this log applies. On the submitted model that means reducing the net additional cost by about two-fifths, from about £3.1m to about £1.9m a year. At the submitted price the medicine displaces more health than it delivers, about 48 QALYs a year net, and the patients who lose that health are unidentifiable, which is a reason to weigh them carefully rather than a reason to discount them. The restricted option is not recommended despite its more favourable ratio: that ratio comes from a subgroup analysis the submission does not adequately support, its net health effect is about zero even on that analysis, and prescribing has previously drifted beyond stated restrictions undetected.